ENDO KAZUKI
| Faculty of Humanities and Social Sciences,Department of Business,Management Information Course | Professor |
| Graduate School of Information Sciences,Major of Information Sciences | Professor |
Last Updated :2025/11/06
■Researcher basic information
Field Of Study
■Career
Career
- Apr. 2025 - Present
Teikyo Heisei University, Graduate School of Information Sciences, Professor - Apr. 2025 - Present
Teikyo Heisei University, Management Information Course, Department of Business, Faculty of Humanities and Social Sciences, Professor - Apr. 2024 - Mar. 2025
Teikyo Heisei University, Graduate School of Environmental Informations, Associate Professor - Apr. 2022 - Mar. 2025
Teikyo Heisei University, Management Information Course, Department of Business, Faculty of Humanities and Social, Associate Professor - Nov. 2014 - Mar. 2022
Topology Inc. - Jul. 2013 - Oct. 2014
Mizuho Bank Co., Ltd., Market Coordination Division - Apr. 2002 - Jun. 2013
Mizuho Corporate Bank Co., Ltd. - Apr. 1999 - Mar. 2002
Industrial Bank of Japan Co., Ltd.
Educational Background
- Apr. 2019 - Mar. 2022, Tokyo Institute of Technology, School of Engineering, Department of Systems and Control Engineering, Doctoral Program
- Apr. 1997 - Mar. 1999, Tokyo Institute of Technology, Graduate School of Decision Science and Technology, Department of Industrial Engineering and Management, Master Course
- Apr. 1993 - Mar. 1997, Tokyo Institute of Technology, School of Science, Mathematics
■Research activity information
Paper
- セルフストレージ投資の損失リスクに備えた準備資金額の計算方法
Kazuki ENDO
Communications of the Operations Research Society of Japan, Sep. 2025, [Reviewed]
This paper proposes a method for calculating the capital required to prepare for losses associated with self-storage investments by using VaR under stochastic models. Here, we assumed the time horizons of VaR as the time when the expected monthly PL is profitable first, and the accumulated expected PL is profitable first. Moreover, we performed some numerical experiments for an investment example.
(This paper is written in Japanese.) - Degraded Image Classification using Knowledge Distillation and Robust Data Augmentations
Dinesh DAULTANI; Masayuki TANAKA; Masatoshi OKUTOMI; Kazuki ENDO
IEICE Transactions on Information and Systems, 01 Dec. 2024, [Reviewed]
This paper proposes an effective combination of data augmentations to train the classification network of degraded images using knowledge distillation.
The results show that our proposed method outperforms existing methods regarding the interval mean accuracy of all degradation levels. - Diffusion-Based Adaptation for Classification of Unknown Degraded Images
Dinesh Daultani; Masayuki Tanaka; Masatoshi Okutomi; Kazuki Endo
New Trends in Image Restoration and Enhancement workshop and challenges (NTIRE2024), CVPR2024 Workshop, Jun. 2024, [Reviewed]
We proposed a knowledge distillation approach to classify degraded images with unknown degradation by using the diffusion-based approach to adapt the degraded images.
There is no classification of writing assignments. - Data augmentation technique for degraded images without losing the classification ability of clean images
Kazuki Endo
Journal of Electronic Imaging, Mar. 2024, [Reviewed]
A classifier trained with degraded images usually underperforms a classifier trained with only clean images in the classification of clean images.
This study proposes a data augmentation technique for degraded images without losing the classification ability of clean images. - ILIAC: Efficient classification of degraded images using knowledge distillation with cutout data augmentation
Dinesh Daultani; Masayuki Tanaka; Masatoshi Okutomi; Kazuki Endo
IS&T Electronic Imaging 2023, Jan. 2023, [Reviewed]
We numerically showed the effectiveness of the cutout data augmentation under the knowledge distillation to train the classifier of degraded images.
There is no classification of writing assignments. - Semantic Segmentation of Degraded Images Using Layer-Wise Feature Adjustor
Kazuki Endo; Masayuki Tanaka; Masatoshi Okutomi
IEEE/CVF Winter Conference on Applications of Computer Vision, Jan. 2023, [Reviewed]
This paper proposes the layer-wise feature adjustor of semantic segmentation for degraded images with various levels of degradation.
The proposed method can recognize not only degraded images but also clean images well. - Determination of Thermal Effusivity of Lunar Regolith Simulant Particle Using Thermal Microscopy
Rie Endo; Yuto Suganuma; Kazuki Endo; Tsuyoshi Nishi; Hiromichi Ohta; Sumitaka Tachikawa
International Journal of Thermophysics, May 2022, [Reviewed] - Image Recognition Network of Several Levels of Degradation (doctoral thesis)
Kazuki Endo
Graduate Major in Systems and Control Engineering, Department of Systems and Control Engineering, School of Engineering, Tokyo Institute of Technology, Mar. 2022 - CNN-Based Classification of Degraded Images with Awareness of Degradation Levels
Kazuki Endo; Masayuki Tanaka; Masatoshi Okutomi
IEEE Transactions on Circuits and Systems for Video Technology, Oct. 2021, [Reviewed]
This paper is an extended version of "Classifying Degraded Images Over Various Levels of Degradation." This paper improved the classification ability of degraded images by using ensemble weights depending on degradation levels and features of restored images.
This paper was seen as an early access paper in December 2020. - CNN-Based Classification of Degraded Images Without Sacrificing Clean Images
Kazuki Endo; Masayuki Tanaka; Masatoshi Okutomi
IEEE Access, Aug. 2021, [Reviewed]
This paper proposed multi-task learning for the classification of degraded images and the estimation of degradation levels by using consistency regularization for image features. The results showed that the proposed method was able to classify degraded images without sacrificing the classification performance of clean images. - スマホで試せる深層学習を用いた物体検出
Jun. 2021, [Reviewed] - 多様な劣化水準に対応可能な劣化画像のクラス分類ネットワーク
Jun. 2021, [Reviewed] - Classifying Degraded Images Over Various Levels of Degradation
Kazuki Endo; Masayuki Tanaka; Masatoshi Okutomi
IEEE International Conference on Image Processing 2020, Oct. 2020, [Reviewed]
This paper proposed the ensemble network of a classification network trained with clean images and one trained with restored images where ensemble weights of the ensemble network depend on estimated degradation levels. - 畳み込みニューラルネットワークを用いた劣化画像のクラス分類
Jun. 2020, [Reviewed] - CNN-based Classification of Degraded Images
Kazuki Endo; Masayuki Tanaka; Masatoshi Okutomi
IS&T International Symposium on Electronic Imaging 2020, Jan. 2020, [Reviewed]
This paper proposed a convolutional neural network whose inputs are both a degraded image and a degradation level. The proposed network improved the classification accuracy against existing methods. - Optimal consumption and portfolio problem with path dependent utility function (master thesis)
Kazuki Endo
Department of Industrial Engineering and Management, Graduate School of Decision Science and Technology, Tokyo Institute of Technology, Mar. 1999
Books and other publications
Lectures, oral presentations, etc.
- Mathematical modelling on self-storage investments
Kazuki ENDO
The 2023 Spring National Conference of Operations Research of Japan, 08 Mar. 2023
We presented the stochastic model of self-storage investments in the Japanese self-storage market. - Semantic Segmentation of Degraded Images Using Layer-Wise Feature Adjustor
Kazuki Endo; Masayuki Tanaka; Masatoshi Okutomi
IEEE/CVF Winter Conference on Applications of Computer Vision, Jan. 2023 - 多様な劣化水準に対応可能な劣化画像のクラス分類ネットワーク
Jun. 2021 - Classifying Degraded Images Over Various Levels of Degradation
Kazuki Endo; Masayuki Tanaka; Masatoshi Okutomi
IEEE International Conference on Image Processing 2020, Oct. 2020 - 畳み込みニューラルネットワークを用いた劣化画像のクラス分類
Jun. 2020 - CNN-based Classification of Degraded Images
Kazuki Endo; Masayuki Tanaka; Masatoshi Okutomi
IS&T International Symposium on Electronic Imaging, Jan. 2020
Affiliated academic society
Research Themes
- 多様な劣化水準に対応可能な劣化画像認識ネットワークに対するデータ拡張方法の研究
Aug. 2022 - Mar. 2025
■University education and qualification information
Qualifications, licenses
- Applied Information Technology Engineer Examination
- Passed a class II information technology engineer examination.
- Passed systems administrator examination
- Certificate-holder Member of the Securities Analysts Association of Japan
- Real estate broker
- 2nd grade Certified Skilled Professional of Financial Planning
- Type-1 High school teaching license for mathematics
- Type-1 Junior high school teaching license for mathematics
- Passed the national examination for assistant
